Why ATS Scores Disagree: One CV, Wildly Different Verdicts

Run the same CV through several ATS checkers and you will often get wildly different scores: one tool says you are in great shape, another flags you as a near-fail. Same document, same job description, different verdicts. If you have ever uploaded your CV to one of these tools and panicked at a 50/100, you are not alone, and the score is probably less meaningful than the panic suggests.

Here is why the tools disagree, and what to trust when the numbers do not line up.

There is no standard ATS score

The biggest reason the numbers do not line up is that each tool is measuring a slightly different thing. There is no industry-standard formula for an "ATS score." Each company invented its own rubric, and the rubrics weight very different signals:

• Keyword density: how many of the job description's exact terms appear in your CV. Some tools lean almost entirely on this.
• Semantic match: whether your CV means the same thing as the JD, even with different words.
• Structural / parseability checks: is your file actually readable by a real ATS? Most tools test this. The signal is binary: it parses or it does not.
• "Impact" scoring: how many of your bullets quantify outcomes ("increased revenue by 27%" vs "managed sales pipeline"). Some tools weight this heavily.
• Section presence: do you have a Skills section, an Education section, a Summary?

A CV that scores high on a tool that mostly checks parseability and keywords can score low on a tool that punishes you for every unquantified bullet. Neither tool is wrong; they are answering different questions.

Free tiers can run a different model

Many checkers, atspass.com included, run a lighter, cheaper AI model on their free tier and reserve the heavier analysis for paying users. That is not inherently dishonest; what matters is whether the scoring rubric stays calibrated the same way. If a free score is generated by a meaningfully different algorithm than the paid one, it becomes an unreliable comparator across tools. Worth keeping in mind when two free scores disagree.

A score is partly a marketing surface

This is the uncomfortable one. A 42/100 generates anxiety. A 78/100 makes you feel done. Anxiety converts. It is a well-documented SaaS pattern to show a user a worrying number and then offer to fix it for a monthly subscription. That does not mean any given score is fabricated, but it does mean a score can be partly a marketing lever, not a pure measurement. If a tool's first action after showing you a low score is to ask for your card, treat the number with appropriate scepticism.

Tools assume different ideal CVs

Some tools assume you are applying for a US corporate role. Others expect a UK-style CV. Some prefer concise summaries; others penalise their absence. A creative-leaning CV that wins recruiter attention can score badly on a tool that expects a strict corporate template, and vice versa.

What you should actually trust

1. Trust structural feedback over score numbers

"Your CV is in PDF format and parses correctly" is verifiable. "Your score is 62" is opaque. If a tool flags a real structural problem, such as tables that confuse the parser, images blocking text, or unusual fonts, fix it. That advice is grounded.

2. Trust specific feedback over vague feedback

"Add the keyword stakeholder management" is concrete and testable. "Improve clarity and impact" is not. Tools that tell you specifically what to change are more useful than tools that give you a number and a vibe.

3. Trust the consensus of issues, not the consensus of scores

If three tools all flag your missing "cross-functional collaboration" keyword, that is a real signal. If three tools give you a different overall number, that is just rubric variance.

4. Run two tools, not five

Two well-chosen tools will catch most of the structural issues and keyword gaps. Five is overkill and the contradictions will paralyse you. Pick one for parseability (most tools do this fine) and one for keyword/semantic depth.

The honest take on atspass.com

I run atspass.com, so take the rest with the appropriate grain of salt. Here is what we deliberately do and do not do:

• We do not lower your free-tier score to push a sale. Free tier uses a smaller AI model, but the rubric is calibrated the same as the paid tier. We would rather you stay free than feel tricked.
• We do not store your CV. Processed in memory, discarded after the response. The score does not get logged alongside your file.
• We measure: keyword match, ATS parseability, structure, quantified achievements, section completeness. The free tier shows you the score, the matched and missing keywords, three rewrite suggestions, and category breakdowns. Paid unlocks all of the rewrites and a downloadable PDF.
• We do not measure visual design. Colour palettes, fonts beyond parseability, infographics: none of that is in our rubric.

If you are sceptical of any of this, run your CV through atspass.com for free (no sign-up, three checks per day) and compare the score to whatever you got elsewhere. The point is not that we are right and everyone else is wrong; the point is that you should not take any single score as gospel.

Four CV fixes that improve every score on every tool

Forget the rubric divergence for a minute. These four fixes will improve your CV on every ATS checker that exists, because they target the underlying signals all tools care about:

1. Quantify two-thirds of your bullets. Not all of them, not none of them, just about two-thirds. Replace "managed a sales pipeline" with "managed a £4M sales pipeline across 12 enterprise accounts."
2. Mirror exact JD phrasing in three to five places. If the JD says "cross-functional collaboration," use that exact phrase. If it says "OKRs," use OKRs.
3. Use standard section headers. Experience, Skills, Education. Avoid creative section names like "What I'm About" or "My Journey."
4. Export to PDF, single-column, standard fonts. No tables, no text boxes, no graphics covering text. The fancier your layout, the more likely the ATS misreads it.

Apply those four and re-run your CV through any tool. The score will go up. If it does not, the rubric is broken, not your CV.

The bottom line

ATS scores are a useful directional signal, not a verdict. The score on its own tells you nothing; the specific feedback attached to it tells you everything. Trust verifiable structural flags, mirror real JD phrasing, quantify outcomes, and stop paying a monthly subscription for a number generated by a rubric you cannot see.

If you want to sanity-check a score from another tool, you can run your CV through atspass.com right now: no sign-up, no card, three free analyses per day. Then ignore whichever score bothers you more.